Re: Version 3.32 of Lout now available

[Jeff Kingston]

The following message is a reply to some email to me from Ludovic
Courtes in which he profiled a Lout run to see whether there was
anything that could be easily accelerated, particularly when the
new -r flag is used.

CopyObject is basically about expanding definitions when they are
invoked.  DisposeObject is about recycling memory at the end of
each page.  I think you have just proved that it would be no use
to optimize the things we talked about, since all the stuff at
the top except SearchSym has to be done on each run.

Lout does not build a parse tree and then build other structures.
It builds a parse tree and then mutates it into the structure that
can be traversed for printing.  At the end of one run there is
currently nothing usable by the next run.  Changing that would
be a terrible job.

The only effective way to save time would be to identify subexpressions
that are not affected by cross referencing and keep them around from
one run to the next.  Again, a huge piece of non-trivial work.

Jeff